Search Results

Search found 11531 results on 462 pages for 'cpu cache'.

Page 128/462 | < Previous Page | 124 125 126 127 128 129 130 131 132 133 134 135  | Next Page >

• iPhone: Low memory crash...

  - by MacTouch

  Once again I'm hunting memory leaks and other crazy mistakes in my code. :) I have a cache with frequently used files (images, data records etc. with a TTL of about one week and a size limited cache (100MB)). There are sometimes more then 15000 files in a directory. On application exit the cache writes an control file with the current cache size along with other useful information. If the applications crashes for some reason (sh.. happens sometimes) I have in such case to calculate the size of all files on application start to make sure I know the cache size. My app crashes at this point because of low memory and I have no clue why. Memory leak detector does not show any leaks at all. I do not see any too. What's wrong with the code below? Is there any other fast way to calculate the total size of all files within a directory on iPhone? Maybe without to enumerate the whole contents of the directory? The code is executed on the main thread. NSUInteger result = 0; NSAutoreleasePool *pool = [[NSAutoreleasePool alloc] init]; NSDirectoryEnumerator *dirEnum = [[[NSFileManager defaultManager] enumeratorAtPath:path] retain]; int i = 0; while ([dirEnum nextObject]) { NSDictionary *attributes = [dirEnum fileAttributes]; NSNumber* fileSize = [attributes objectForKey:NSFileSize]; result += [fileSize unsignedIntValue]; if (++i % 500 == 0) { // I tried lower values too [pool drain]; } } [dirEnum release]; dirEnum = nil; [pool release]; pool = nil; Thanks, MacTouch

  Read the article

• Ajax Request using jQuery in Rails

  - by Steve

  Hi... I am sending an Ajax Request using jQuery. What happens is that I am getting an "405 Method Not Allowed" Error. I am just posting a form, which would get the detail from the form and insert it into the DB. Just the usual stuff.I am using WEBrick that comes as default with the rails package. Can somebody please tell me how to fix this. This is the code that triggers the Ajax Request $.post($(this).attr("action") + ".js",$(this).serialize(),null,"script"); Response Headers Cache-Control no-cache Allow GET, PUT, DELETE Content-Type text/html; charset=utf-8 Content-Length 9502 Server WEBrick/1.3.1 (Ruby/1.9.1/2009-12-07) Date Wed, 02 Jun 2010 20:41:33 GMT Connection Keep-Alive Request Headers Host localhost:3000 User-Agent Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.6; en-US; rv:1.9.2.3) Gecko/20100401 Firefox/3.6.3 Accept application/json, text/javascript, */* Accept-Language en-us,en;q=0.5 Accept-Encoding gzip,deflate Accept-Charset ISO-8859-1,utf-8;q=0.7,*;q=0.7 Keep-Alive 115 Connection keep-alive Content-Type application/x-www-form-urlencoded; charset=UTF-8 X-Requested-With XMLHttpRequest Referer http://localhost:3000/viewspot/3 Content-Length 141 Pragma no-cache Cache-Control no-cache

  Read the article

• ASP.net web page still displaying cached versions

  - by user279521

  My web page is still displaying a previously cached versions of the page. I have this in the page_load event: Response.Clear(); Response.Buffer = true; Response.ExpiresAbsolute = DateTime.Now.AddDays(-1d); Response.Expires = -1; Response.CacheControl = "no-cache"; Response.Cache.SetCacheability(HttpCacheability.NoCache); I have this in the Page_Init: protected void Page_Init(object Sender, EventArgs e) { Response.Cache.SetCacheability(HttpCacheability.NoCache); Response.Cache.SetExpires(DateTime.Now.AddDays(-1)); } Any idea what I might be missing?

  Read the article

• Is a Multi-DAL Approach the way to go here?

  - by Krisc

  Working on the data access / model layer in this little MVC2 project and trying to think things out to future projects. I have a database with some basic tables and I have classes in the model layer that represent them. I obviously need something to connect the two. The easiest is to provide some sort of 'provider' that can run operations on the database and return objects. But this is for a website that would potentially be used "a lot" (I know, very general) so I want to cache results from the data layer and keep the cache updated as new data is generated. This question deals with how best to approach this problem of dual DALS. One that returns cached data when possible and goes to the data layer when there is a cache miss. But more importantly, how to integrate the core provider (thing that goes into database) with the caching layer so that it too can rely on cached objects rather than creating new ones. Right now I have the following interfaces: IDataProvider is used to reach the database. It doesn't concern itself with the meaning of the objects it produces, but simply the way to produce them. interface IDataProvider{ // Select, Update, Create, et cetera access IEnumerable<Entry> GetEntries(); Entry GetEntryById(int id); } IDataManager is a layer that sits on top of the IDataProvider layer and manages the cache interface IDataManager : IDataProvider{ void ClearCache(); } Note that in practice the IDataManager implementation will have useful helper functions to add objects to their related cache stores. (In the future I may define other functions on the interface) I guess what I am looking for is the best way to approach a loop back from the IDataProvider implementations so that they can access the cache. Or a different approach entirely may be in order? I am not very interested in 3rd party products at the moment as I am interested in the design of these things much more than this specific implementation. Edit: I realize the title may be a bit misleading. I apologize for that... not sure what to call this question.

  Read the article

• Will MySql caching cause performance problems?

  - by Camran

  I am about to upload my website onto a VPS. It is a classifieds website, where all data is stored in MySql and Solr. I wonder if when using MySql:s cache, the server will slow down? Ie, if somebody makes a search for the first time, and MySql is to cache the query, will the caching make the server slower than if it would not cache anything? After the caching is done I know things will improve in terms of performance... But I would like to know if I should even use the cache or not, what do you think? Thanks

  Read the article

• JavaScript doesn't parse when mod-rewrited through a PHP file?

  - by Newbtophp

  If I do the following (this is the actual/direct path to the JavaScript file): <script href="http://localhost/tpl/blue/js/functions.js" type="text/javascript"></script> It works fine, and the JavaScript parses - as its meant too. However I'm wanting to shorten the path to the JavaScript file (aswell as do some caching) which is why I'm rewriting all JavaScript files via .htaccess to cache.php (which handles the caching). The .htaccess contains the following: <IfModule mod_rewrite.c> RewriteEngine On RewriteBase / RewriteRule ^js/(.+?\.js)$ cache.php?file=$1 [NC] </IfModule> cache.php contains the following PHP code: <?php if (extension_loaded('zlib')) { ob_start('ob_gzhandler'); } $file = basename($_GET['file']); if (file_exists("tpl/blue/js/".$file)) { header("Content-Type: application/javascript"); header('Cache-Control: must-revalidate'); header('Expires: ' . gmdate('D, d M Y H:i:s', time() + 3600) . ' GMT'); echo file_get_contents("tpl/blue/js/".$file); } ?> and I'm calling the JavaScript file like so: <script href="http://localhost/js/functions.js" type="text/javascript"></script> But doing that the JavaScript doesn't parse? (if I call the functions which are within functions.js later on in the page they don't work) - so theirs a problem either with cache.php or the rewrite rule? (because the file by itself works fine). If I access the rewrited file- http://localhost/js/functions.js directly it prints the JavaScript code, as any JavaScript file would - so I'm confused as to what I'm doing wrong... All help is appreciated! :)

  Read the article

• How to trigger code on two different servers in a WAS cluster?

  - by Dean J

  I have an administrative page in a web application that resets the cache, but it only resets the cache on the current JVM. The web application is deployed as a cluster to two WAS servers. Any way that I can elegantly have the "clear cache" button on each server trigger the method on both JVMs? Edit: The original developer just wrote a singleton holding a HashMap to be the cache in question. Lightweight and (previously) worked just fine for the requirements. It caches content pulled from six or seven web services for specified amounts of time. Edit: The entire application in question is three pages, so the elegant solution might well be the lightest solution.

  Read the article

• Chain of Responsibility Pattern: is it a good practice to have interdependent handlers?

  - by wei

  I have this scenario: I have a chain of query handlers, the first is to query the cache, if the cache can't answer it or the answer is stale, then hit a database, if it can't find the answer or the answer is stale again, then query a remote web service. But I am not sure if this is the right way to use this pattern, since the work flow is pretty much fixed, and the cache and database handlers depend on the next step's return result to refresh its records.

  Read the article

• Problem accessing private variables in jQuery like chainable design pattern

  - by novogeek

  Hi folks, I'm trying to create my custom toolbox which imitates jQuery's design pattern. Basically, the idea is somewhat derived from this post: http://stackoverflow.com/questions/2061501/jquery-plugin-design-pattern-common-practice-for-dealing-with-private-function (Check the answer given by "David"). So here is my toolbox function: (function(window){ var mySpace=function(){ return new PrivateSpace(); } var PrivateSpace=function(){ var testCache={}; }; PrivateSpace.prototype={ init:function(){ console.log('init this:', this); return this; }, ajax:function(){ console.log('make ajax calls here'); return this; }, cache:function(key,selector){ console.log('cache selectors here'); testCache[key]=selector; console.log('cached selector: ',testCache); return this; } } window.hmis=window.m$=mySpace(); })(window) Now, if I execute this function like: console.log(m$.cache('firstname','#FirstNameTextbox')); I get an error 'testCache' is not defined. I'm not able to access the variable "testCache" inside my cache function of the prototype. How should I access it? Basically, what I want to do is, I want to cache all my jQuery selectors into an object and use this object in the future.

  Read the article

• Coldfusion 8 and HTTP PUT - is there a way to PUT an object?

  - by ciaranarcher

  Hi all We are using EHCache with CF 8 to cache stuff on a central server using a RESTful interface over HTTP. I am trying to cache a cfquery object to the cache server. I can get this to work if I call EHCache direct (i.e. store it in a local cache) but if I try to cache on a remote server over HTTP I am running into problems. The code I am using is as follows: <cfhttp url="http://localhost:8080/myCache/myKey" method="put" result="r" timeout="2" throwonerror="true" > <cfhttpparam type="body" value="#ARGUMENTS.item#" /> </cfhttp> CF doesn't like this reference to #ARGUMENTS.item# and it complains Complex object types cannot be converted to simple values. Can anyone give me an example of how to put an object over http using CF? If this is not possible with CF then a Java example would be the next best thing. Many thanks in advance! PS: I do not want to use serialization to text/JSON etc. as this approach has issues with data integrity and most importantly it's not fast enough.

  Read the article

• How do I ensure jQuery ajax call does not send a local copy of file?

  - by Tommy

  $.ajax({ type: 'GET', url: "string.txt", cache: false, success: function(str){ alert("Data is: "+ str); } }); In this example, string.txt is still sent to the cache (\Temporary Internet Files) How do I ensure that the file is not sent - read from the server only? Am I missing an option? I set cache to false but that does not block it from being sent to client. For example, ajax POST does not send a local copy..... Thanks

  Read the article

• How should I implement simple caches with concurrency on Redis?

  - by solublefish

  Background I have a 2-tier web service - just my app server and an RDBMS. I want to move to a pool of identical app servers behind a load balancer. I currently cache a bunch of objects in-process. I hope to move them to a shared Redis. I have a dozen or so caches of simple, small-sized business objects. For example, I have a set of Foos. Each Foo has a unique FooId and an OwnerId. One "owner" may own multiple Foos. In a traditional RDBMS this is just a table with an index on the PK FooId and one on OwnerId. I'm caching this in one process simply: Dictionary<int,Foo> _cacheFooById; Dictionary<int,HashSet<int>> _indexFooIdsByOwnerId; Reads come straight from here, and writes go here and to the RDBMS. I usually have this invariant: "For a given group [say by OwnerId], the whole group is in cache or none of it is." So when I cache miss on a Foo, I pull that Foo and all the owner's other Foos from the RDBMS. Updates make sure to keep the index up to date and respect the invariant. When an owner calls GetMyFoos I never have to worry that some are cached and some aren't. What I did already The first/simplest answer seems to be to use plain ol' SET and GET with a composite key and json value: SET( "ServiceCache:Foo:" + theFoo.Id, JsonSerialize(theFoo)); I later decided I liked: HSET( "ServiceCache:Foo", theFoo.FooId, JsonSerialize(theFoo)); That lets me get all the values in one cache as HVALS. It also felt right - I'm literally moving hashtables to Redis, so perhaps my top-level items should be hashes. This works to first order. If my high-level code is like: UpdateCache(myFoo); AddToIndex(myFoo); That translates into: HSET ("ServiceCache:Foo", theFoo.FooId, JsonSerialize(theFoo)); var myFoos = JsonDeserialize( HGET ("ServiceCache:FooIndex", theFoo.OwnerId) ); myFoos.Add(theFoo.OwnerId); HSET ("ServiceCache:FooIndex", theFoo.OwnerId, JsonSerialize(myFoos)); However, this is broken in two ways. Two concurrent operations can read/modify/write at the same time. The latter "wins" the final HSET and the former's index update is lost. Another operation could read the index in between the first and second lines. It would miss a Foo that it should find. So how do I index properly? I think I could use a Redis set instead of a json-encoded value for the index. That would solve part of the problem since the "add-to-index-if-not-already-present" would be atomic. I also read about using MULTI as a "transaction" but it doesn't seem like it does what I want. Am I right that I can't really MULTI; HGET; {update}; HSET; EXEC since it doesn't even do the HGET before I issue the EXEC? I also read about using WATCH and MULTI for optimistic concurrency, then retrying on failure. But WATCH only works on top-level keys. So it's back to SET/GET instead of HSET/HGET. And now I need a new index-like-thing to support getting all the values in a given cache. If I understand it right, I can combine all these things to do the job. Something like: while(!succeeded) { WATCH( "ServiceCache:Foo:" + theFoo.FooId ); WATCH( "ServiceCache:FooIndexByOwner:" + theFoo.OwnerId ); WATCH( "ServiceCache:FooIndexAll" ); MULTI(); SET ("ServiceCache:Foo:" + theFoo.FooId, JsonSerialize(theFoo)); SADD ("ServiceCache:FooIndexByOwner:" + theFoo.OwnerId, theFoo.FooId); SADD ("ServiceCache:FooIndexAll", theFoo.FooId); EXEC(); //TODO somehow set succeeded properly } Finally I'd have to translate this pseudocode into real code depending how my client library uses WATCH/MULTI/EXEC; it looks like they need some sort of context to hook them together. All in all this seems like a lot of complexity for what has to be a very common case; I can't help but think there's a better, smarter, Redis-ish way to do things that I'm just not seeing. How do I lock properly? Even if I had no indexes, there's still a (probably rare) race condition. A: HGET - cache miss B: HGET - cache miss A: SELECT B: SELECT A: HSET C: HGET - cache hit C: UPDATE C: HSET B: HSET ** this is stale data that's clobbering C's update. Note that C could just be a really-fast A. Again I think WATCH, MULTI, retry would work, but... ick. I know in some places people use special Redis keys as locks for other objects. Is that a reasonable approach here? Should those be top-level keys like ServiceCache:FooLocks:{Id} or ServiceCache:Locks:Foo:{Id}? Or make a separate hash for them - ServiceCache:Locks with subkeys Foo:{Id}, or ServiceCache:Locks:Foo with subkeys {Id} ? How would I work around abandoned locks, say if a transaction (or a whole server) crashes while "holding" the lock?

  Read the article

• Execute a command using php under ssh2 in php

  - by Mervyn

  Using Mint terminal my script connects using ssh2_connect and ssh2_auth-password. When am logged in successfully I want to run a command which will give me the hardware cpu. Is there a way I can use to exec the command in my script then show the results. I have used system and exec for pinging. if i was in the terminal i do the login. then type "get hardware cpu" in the terminal it would look like this: Test~ $ get hardware cpu

  Read the article

• NoMethodError using Memcached Sinatra-style

  - by sevennineteen

  I've just installed Memcached on my Mac and updated my Sinatra app configuration as described in Heroku's documentation, but I'm getting a NoMethodError when trying to use the hash-based syntax they specify: >> CACHE['color'] = 'blue' >> CACHE['color'] Using explicit get and set methods as below seems to work fine. >> CACHE.set('color', 'blue') >> CACHE.get('color') If necessary I can use the latter syntax, but the former seems more elegant. I haven't tested this on Heroku's environment since I'd like whatever implementation I use to work on my local environment as well. Thanks!

  Read the article

• Why changing the images name on server results in calling the old ones?

  - by moderns

  I am running a slideshow on Ubuntu 12.04.1 that loads the images (slide1.jpg, slide2.jpg, slide3.jpg.., slide5.jpg) using the Javascript and styles as below: document.getElementById('slide_area').className='slide'+step; .slide1{background-image: url(../upload/slide1.jpg)} .slide2{background-image: url(../upload/slide2.jpg)} .slide3{background-image: url(../upload/slide3.jpg)} .slide4{background-image: url(../upload/slide4.jpg)} .slide5{background-image: url(../upload/slide5.jpg)} When I change the images names (show1.jpg, show2.jpg, show3.jpg.., show5.jpg) and also change the style as below: .slide1{background-image: url(../upload/show1.jpg)} .slide2{background-image: url(../upload/show2.jpg)} .slide3{background-image: url(../upload/show3.jpg)} .slide4{background-image: url(../upload/show4.jpg)} .slide5{background-image: url(../upload/show5.jpg)} And open the network section on Chrome, I see the server is calling the new name and old name for images! I added the header in the index.php: header("Cache-Control: no-store, no-cache, must-revalidate, max-age=0"); header("Expires: Sat, 26 Jul 1997 05:00:00 GMT"); // Date in the past header("Cache-Control: post-check=0, pre-check=0", false); header("Pragma: no-cache"); Nothing worked out with me and the slideshow doesn't work properly when I change the name of images even when clearing the browser cache as I load images sequentially (one by one) depending on imageObject.complete property! But without changing the name everything is going perfect and the images are loaded smoothly! Thank you for your help!

  Read the article

• Caching in Ruby Gem, possibly not using Rails

  - by corprew

  I am rewriting an existing Ruby Gem to include caching. This is for a gem that is relatively commonly used, and accesses a large amount of static data on a web service. Currently, I have a small number of gem users doing a large number of accesses to the service that under normal conditions would be swamping / downing the service, and we're going to put the gem up on github for general consumption. Right now, users can choose between using the rails cache mechanism, a simple disk cache, or no cache. What is best practice for letting people choose what cache to use like this (being able to use this outside of rails is a priority so i can't just bail to the underlying caching mechanism)? I'm looking for suggestions/examples for configuration and interface, especially. Thanks for your suggestions

  Read the article

• php mp3 headers in google chrome

  - by David

  I have this in a php to show a mp3 file, it code works fine on firefox and explorer but in chrome it not work. The chrome player appears but no sound and not increases time $ext = strtolower(substr(strrchr($filename,"."),1)); $ctype="audio/mpeg"; header( 'Expires: Sat, 26 Jul 1997 05:00:00 GMT' ); header( 'Last-Modified: ' . gmdate( 'D, d M Y H:i:s' ) . ' GMT' ); header( 'Cache-Control: no-store, no-cache, must-revalidate' ); header( 'Cache-Control: post-check=0, pre-check=0', false ); header( 'Pragma: no-cache' ); header("Content-Type: $ctype"); header("Content-Length: ".$len);

  Read the article

• Extract information from a Func<bool, T> or alike lambda

  - by Syska

  I''m trying to build a generic cache layer. ICacheRepository Say I have the following: public class Person { public int PersonId { get; set; } public string Firstname { get; set; } public string Lastname { get; set; } public DateTime Added { get; set; } } And I have something like this: list.Where(x => x.Firstname == "Syska"); Here I want to extract the above information, to see if the query supplied the "PersonId" which it did not, so I dont want to cache it. But lets say I run a query like this: list.Where(x => x.PersonId == 10); Since PersonId is my key ... I want to cache it. with the key like "Person_10" and I later can fetch it from the cache. I know its possible to extract the information with Expression<Func<>> but there seems to be a big overhead of doing this (when running compile and extract the Constant values etc. and a bunch of cache to be sure to parse right) Are there a framework for this? Or some smart/golden way of doing this ?

  Read the article

• test php disabled caching

  - by user1691389

  My site had a problem in that certain browsers (especially opera and gecko) were "over-caching" (caching far too much for my taste). I've just added the following PHP snippet to hopefully disable caching in all browsers: <?php header("Expires: Tue, 01 Jan 2000 00:00:00 GMT"); header("Last-Modified: " . gmdate("D, d M Y H:i:s") . " GMT"); header("Cache-Control: no-store, no-cache, must-revalidate, max-age=0"); header("Cache-Control: post-check=0, pre-check=0", false); header("Pragma: no-cache"); ?> Question: How would you test this out, to make sure it actually works?

  Read the article

• Magento get loaded collection

  - by Zertalx

  Hey guys I got an issue in magento when listing products. $_productCollection = array(); $key = 'product_collection_'.Mage::app()->getRequest()->getRequestUri(); $_productCollection = Mage::helper('cache')->getDataInCache($key); if (empty($_productCollection)) { $_productCollection=$this->getLoadedProductCollection(); Mage::helper('cache')->setDataInCache($key,$_productCollection,'front_end_collection'); } That's my code, i'm loading the collection and saved in cache, the problem is that I added new products from the admin panel but they are not showing up in the listing. I have refreshed cache, reindex all the data and checke that the products were in stock. Also check the visibility because i understand that the collection gets just the visible products, am i correct? I'm using magento enterprise 1.9 I'm new in magento and i'm having troubles understanding how the Collections works, i hope somebody can help me. Thanks

  Read the article

• Java: "cannot find symbol" error of a String[] defined within a while-loop

  - by David

  Here's the relevant code: public static String[] runTeams (String CPUcolor) { boolean z = false ; //String[] a = new String[6] ; boolean CPU = false ; while (z == false) { while (CPU==false) { String[] a = assignTeams () ; printOrder (a) ; for (int i = 1; i<a.length; i++) { if (a[i].equals(CPUcolor)) CPU = true ; } if (CPU==false) { System.out.println ("ERROR YOU NEED TO INCLUDE THE COLOR OF THE CPU IN THE TURN ORDER") ; } } System.out.println ("is this turn order correct? (Y/N)") ; String s = getIns () ; while (!((s.equals ("y")) || (s.equals ("Y")) || (s.equals ("n")) || (s.equals ("N")))) { System.out.println ("try again") ; s = getIns () ; } if (s.equals ("y") || s.equals ("Y") ) z = true ; } return a ; } the error i get is: Risk.java:416: cannot find symbol symbol : variable a location: class Risk return a ; ^ Why did i get this error? It seems that a is clearly defined in the line String[] a = assignTeams () ; and if anything is used by the lineprintOrder (a) ;` it seems to me that if the symbol a really couldn't be found then the compiler should blow up there and not at the return statment. (also the method assignTeams returns an array of Strings.)

  Read the article

• Want to disable flash caching ?

  - by Mina Samy

  Hi all I have a .swf flash gallery that loads pics from a XML file the probelm is when I modify the XML the modifications do not reflect on the flash till I delete the browsing cache from the browser I tried to disable caching using code like this Response.Cache.SetCacheability(HttpCacheability.NoCache); Response.AddHeader("Pragma","no-cache"); Response.Expires = -1; but not working is there any workaround for this ? thanks

  Read the article

• array and array_view from amp.h

  - by Daniel Moth

  This is a very long post, but it also covers what are probably the classes (well, array_view at least) that you will use the most with C++ AMP, so I hope you enjoy it! Overview The concurrency::array and concurrency::array_view template classes represent multi-dimensional data of type T, of N dimensions, specified at compile time (and you can later access the number of dimensions via the rank property). If N is not specified, it is assumed that it is 1 (i.e. single-dimensional case). They are rectangular (not jagged). The difference between them is that array is a container of data, whereas array_view is a wrapper of a container of data. So in that respect, array behaves like an STL container, whereas the closest thing an array_view behaves like is an STL iterator (albeit with random access and allowing you to view more than one element at a time!). The data in the array (whether provided at creation time or added later) resides on an accelerator (which is specified at creation time either explicitly by the developer, or set to the default accelerator at creation time by the runtime) and is laid out contiguously in memory. The data provided to the array_view is not stored by/in the array_view, because the array_view is simply a view over the real source (which can reside on the CPU or other accelerator). The underlying data is copied on demand to wherever the array_view is accessed. Elements which differ by one in the least significant dimension of the array_view are adjacent in memory. array objects must be captured by reference into the lambda you pass to the parallel_for_each call, whereas array_view objects must be captured by value (into the lambda you pass to the parallel_for_each call). Creating array and array_view objects and relevant properties You can create array_view objects from other array_view objects of the same rank and element type (shallow copy, also possible via assignment operator) so they point to the same underlying data, and you can also create array_view objects over array objects of the same rank and element type e.g.   array_view<int,3> a(b); // b can be another array or array_view of ints with rank=3 Note: Unlike the constructors above which can be called anywhere, the ones in the rest of this section can only be called from CPU code. You can create array objects from other array objects of the same rank and element type (copy and move constructors) and from other array_view objects, e.g.   array<float,2> a(b); // b can be another array or array_view of floats with rank=2 To create an array from scratch, you need to at least specify an extent object, e.g. array<int,3> a(myExtent);. Note that instead of an explicit extent object, there are convenience overloads when N<=3 so you can specify 1-, 2-, 3- integers (dependent on the array's rank) and thus have the extent created for you under the covers. At any point, you can access the array's extent thought the extent property. The exact same thing applies to array_view (extent as constructor parameters, incl. convenience overloads, and property). While passing only an extent object to create an array is enough (it means that the array will be written to later), it is not enough for the array_view case which must always wrap over some other container (on which it relies for storage space and actual content). So in addition to the extent object (that describes the shape you'd like to be viewing/accessing that data through), to create an array_view from another container (e.g. std::vector) you must pass in the container itself (which must expose .data() and a .size() methods, e.g. like std::array does), e.g.   array_view<int,2> aaa(myExtent, myContainerOfInts); Similarly, you can create an array_view from a raw pointer of data plus an extent object. Back to the array case, to optionally initialize the array with data, you can pass an iterator pointing to the start (and optionally one pointing to the end of the source container) e.g.   array<double,1> a(5, myVector.begin(), myVector.end()); We saw that arrays are bound to an accelerator at creation time, so in case you don’t want the C++ AMP runtime to assign the array to the default accelerator, all array constructors have overloads that let you pass an accelerator_view object, which you can later access via the accelerator_view property. Note that at the point of initializing an array with data, a synchronous copy of the data takes place to the accelerator, and then to copy any data back we'll see that an explicit copy call is required. This does not happen with the array_view where copying is on demand... refresh and synchronize on array_view Note that in the previous section on constructors, unlike the array case, there was no overload that accepted an accelerator_view for array_view. That is because the array_view is simply a wrapper, so the allocation of the data has already taken place before you created the array_view. When you capture an array_view variable in your call to parallel_for_each, the copy of data between the non-CPU accelerator and the CPU takes place on demand (i.e. it is implicit, versus the explicit copy that has to happen with the array). There are some subtleties to the on-demand-copying that we cover next. The assumption when using an array_view is that you will continue to access the data through the array_view, and not through the original underlying source, e.g. the pointer to the data that you passed to the array_view's constructor. So if you modify the data through the array_view on the GPU, the original pointer on the CPU will not "know" that, unless one of two things happen: you access the data through the array_view on the CPU side, i.e. using indexing that we cover below you explicitly call the array_view's synchronize method on the CPU (this also gets called in the array_view's destructor for you) Conversely, if you make a change to the underlying data through the original source (e.g. the pointer), the array_view will not "know" about those changes, unless you call its refresh method. Finally, note that if you create an array_view of const T, then the data is copied to the accelerator on demand, but it does not get copied back, e.g.   array_view<const double, 5> myArrView(…); // myArrView will not get copied back from GPU There is also a similar mechanism to achieve the reverse, i.e. not to copy the data of an array_view to the GPU. copy_to, data, and global copy/copy_async functions Both array and array_view expose two copy_to overloads that allow copying them to another array, or to another array_view, and these operations can also be achieved with assignment (via the = operator overloads). Also both array and array_view expose a data method, to get a raw pointer to the underlying data of the array or array_view, e.g. float* f = myArr.data();. Note that for array_view, this only works when the rank is equal to 1, due to the data only being contiguous in one dimension as covered in the overview section. Finally, there are a bunch of global concurrency::copy functions returning void (and corresponding concurrency::copy_async functions returning a future) that allow copying between arrays and array_views and iterators etc. Just browse intellisense or amp.h directly for the full set. Note that for array, all copying described throughout this post is deep copying, as per other STL container expectations. You can never have two arrays point to the same data. indexing into array and array_view plus projection Reading or writing data elements of an array is only legal when the code executes on the same accelerator as where the array was bound to. In the array_view case, you can read/write on any accelerator, not just the one where the original data resides, and the data gets copied for you on demand. In both cases, the way you read and write individual elements is via indexing as described next. To access (or set the value of) an element, you can index into it by passing it an index object via the subscript operator. Furthermore, if the rank is 3 or less, you can use the function ( ) operator to pass integer values instead of having to use an index object. e.g. array<float,2> arr(someExtent, someIterator); //or array_view<float,2> arr(someExtent, someContainer); index<2> idx(5,4); float f1 = arr[idx]; float f2 = arr(5,4); //f2 ==f1 //and the reverse for assigning, e.g. arr(idx[0], 7) = 6.9; Note that for both array and array_view, regardless of rank, you can also pass a single integer to the subscript operator which results in a projection of the data, and (for both array and array_view) you get back an array_view of rank N-1 (or if the rank was 1, you get back just the element at that location). Not Covered In this already very long post, I am not going to cover three very cool methods (and related overloads) that both array and array_view expose: view_as, section, reinterpret_as. We'll revisit those at some point in the future, probably on the team blog. Comments about this post by Daniel Moth welcome at the original blog.

  Read the article

• 256 Windows Azure Worker Roles, Windows Kinect and a 90's Text-Based Ray-Tracer

  - by Alan Smith

  For a couple of years I have been demoing a simple render farm hosted in Windows Azure using worker roles and the Azure Storage service. At the start of the presentation I deploy an Azure application that uses 16 worker roles to render a 1,500 frame 3D ray-traced animation. At the end of the presentation, when the animation was complete, I would play the animation delete the Azure deployment. The standing joke with the audience was that it was that it was a “$2 demo”, as the compute charges for running the 16 instances for an hour was $1.92, factor in the bandwidth charges and it’s a couple of dollars. The point of the demo is that it highlights one of the great benefits of cloud computing, you pay for what you use, and if you need massive compute power for a short period of time using Windows Azure can work out very cost effective. The “$2 demo” was great for presenting at user groups and conferences in that it could be deployed to Azure, used to render an animation, and then removed in a one hour session. I have always had the idea of doing something a bit more impressive with the demo, and scaling it from a “$2 demo” to a “$30 demo”. The challenge was to create a visually appealing animation in high definition format and keep the demo time down to one hour.  This article will take a run through how I achieved this. Ray Tracing Ray tracing, a technique for generating high quality photorealistic images, gained popularity in the 90’s with companies like Pixar creating feature length computer animations, and also the emergence of shareware text-based ray tracers that could run on a home PC. In order to render a ray traced image, the ray of light that would pass from the view point must be tracked until it intersects with an object. At the intersection, the color, reflectiveness, transparency, and refractive index of the object are used to calculate if the ray will be reflected or refracted. Each pixel may require thousands of calculations to determine what color it will be in the rendered image. Pin-Board Toys Having very little artistic talent and a basic understanding of maths I decided to focus on an animation that could be modeled fairly easily and would look visually impressive. I’ve always liked the pin-board desktop toys that become popular in the 80’s and when I was working as a 3D animator back in the 90’s I always had the idea of creating a 3D ray-traced animation of a pin-board, but never found the energy to do it. Even if I had a go at it, the render time to produce an animation that would look respectable on a 486 would have been measured in months. PolyRay Back in 1995 I landed my first real job, after spending three years being a beach-ski-climbing-paragliding-bum, and was employed to create 3D ray-traced animations for a CD-ROM that school kids would use to learn physics. I had got into the strange and wonderful world of text-based ray tracing, and was using a shareware ray-tracer called PolyRay. PolyRay takes a text file describing a scene as input and, after a few hours processing on a 486, produced a high quality ray-traced image. The following is an example of a basic PolyRay scene file. background Midnight_Blue   static define matte surface { ambient 0.1 diffuse 0.7 } define matte_white texture { matte { color white } } define matte_black texture { matte { color dark_slate_gray } } define position_cylindrical 3 define lookup_sawtooth 1 define light_wood <0.6, 0.24, 0.1> define median_wood <0.3, 0.12, 0.03> define dark_wood <0.05, 0.01, 0.005>     define wooden texture { noise surface { ambient 0.2  diffuse 0.7  specular white, 0.5 microfacet Reitz 10 position_fn position_cylindrical position_scale 1  lookup_fn lookup_sawtooth octaves 1 turbulence 1 color_map( [0.0, 0.2, light_wood, light_wood] [0.2, 0.3, light_wood, median_wood] [0.3, 0.4, median_wood, light_wood] [0.4, 0.7, light_wood, light_wood] [0.7, 0.8, light_wood, median_wood] [0.8, 0.9, median_wood, light_wood] [0.9, 1.0, light_wood, dark_wood]) } } define glass texture { surface { ambient 0 diffuse 0 specular 0.2 reflection white, 0.1 transmission white, 1, 1.5 }} define shiny surface { ambient 0.1 diffuse 0.6 specular white, 0.6 microfacet Phong 7  } define steely_blue texture { shiny { color black } } define chrome texture { surface { color white ambient 0.0 diffuse 0.2 specular 0.4 microfacet Phong 10 reflection 0.8 } }   viewpoint {     from <4.000, -1.000, 1.000> at <0.000, 0.000, 0.000> up <0, 1, 0> angle 60     resolution 640, 480 aspect 1.6 image_format 0 }       light <-10, 30, 20> light <-10, 30, -20>   object { disc <0, -2, 0>, <0, 1, 0>, 30 wooden }   object { sphere <0.000, 0.000, 0.000>, 1.00 chrome } object { cylinder <0.000, 0.000, 0.000>, <0.000, 0.000, -4.000>, 0.50 chrome }   After setting up the background and defining colors and textures, the viewpoint is specified. The “camera” is located at a point in 3D space, and it looks towards another point. The angle, image resolution, and aspect ratio are specified. Two lights are present in the image at defined coordinates. The three objects in the image are a wooden disc to represent a table top, and a sphere and cylinder that intersect to form a pin that will be used for the pin board toy in the final animation. When the image is rendered, the following image is produced. The pins are modeled with a chrome surface, so they reflect the environment around them. Note that the scale of the pin shaft is not correct, this will be fixed later. Modeling the Pin Board The frame of the pin-board is made up of three boxes, and six cylinders, the front box is modeled using a clear, slightly reflective solid, with the same refractive index of glass. The other shapes are modeled as metal. object { box <-5.5, -1.5, 1>, <5.5, 5.5, 1.2> glass } object { box <-5.5, -1.5, -0.04>, <5.5, 5.5, -0.09> steely_blue } object { box <-5.5, -1.5, -0.52>, <5.5, 5.5, -0.59> steely_blue } object { cylinder <-5.2, -1.2, 1.4>, <-5.2, -1.2, -0.74>, 0.2 steely_blue } object { cylinder <5.2, -1.2, 1.4>, <5.2, -1.2, -0.74>, 0.2 steely_blue } object { cylinder <-5.2, 5.2, 1.4>, <-5.2, 5.2, -0.74>, 0.2 steely_blue } object { cylinder <5.2, 5.2, 1.4>, <5.2, 5.2, -0.74>, 0.2 steely_blue } object { cylinder <0, -1.2, 1.4>, <0, -1.2, -0.74>, 0.2 steely_blue } object { cylinder <0, 5.2, 1.4>, <0, 5.2, -0.74>, 0.2 steely_blue }   In order to create the matrix of pins that make up the pin board I used a basic console application with a few nested loops to create two intersecting matrixes of pins, which models the layout used in the pin boards. The resulting image is shown below. The pin board contains 11,481 pins, with the scene file containing 23,709 lines of code. For the complete animation 2,000 scene files will be created, which is over 47 million lines of code. Each pin in the pin-board will slide out a specific distance when an object is pressed into the back of the board. This is easily modeled by setting the Z coordinate of the pin to a specific value. In order to set all of the pins in the pin-board to the correct position, a bitmap image can be used. The position of the pin can be set based on the color of the pixel at the appropriate position in the image. When the Windows Azure logo is used to set the Z coordinate of the pins, the following image is generated. The challenge now was to make a cool animation. The Azure Logo is fine, but it is static. Using a normal video to animate the pins would not work; the colors in the video would not be the same as the depth of the objects from the camera. In order to simulate the pin board accurately a series of frames from a depth camera could be used. Windows Kinect The Kenect controllers for the X-Box 360 and Windows feature a depth camera. The Kinect SDK for Windows provides a programming interface for Kenect, providing easy access for .NET developers to the Kinect sensors. The Kinect Explorer provided with the Kinect SDK is a great starting point for exploring Kinect from a developers perspective. Both the X-Box 360 Kinect and the Windows Kinect will work with the Kinect SDK, the Windows Kinect is required for commercial applications, but the X-Box Kinect can be used for hobby projects. The Windows Kinect has the advantage of providing a mode to allow depth capture with objects closer to the camera, which makes for a more accurate depth image for setting the pin positions. Creating a Depth Field Animation The depth field animation used to set the positions of the pin in the pin board was created using a modified version of the Kinect Explorer sample application. In order to simulate the pin board accurately, a small section of the depth range from the depth sensor will be used. Any part of the object in front of the depth range will result in a white pixel; anything behind the depth range will be black. Within the depth range the pixels in the image will be set to RGB values from 0,0,0 to 255,255,255. A screen shot of the modified Kinect Explorer application is shown below. The Kinect Explorer sample application was modified to include slider controls that are used to set the depth range that forms the image from the depth stream. This allows the fine tuning of the depth image that is required for simulating the position of the pins in the pin board. The Kinect Explorer was also modified to record a series of images from the depth camera and save them as a sequence JPEG files that will be used to animate the pins in the animation the Start and Stop buttons are used to start and stop the image recording. En example of one of the depth images is shown below. Once a series of 2,000 depth images has been captured, the task of creating the animation can begin. Rendering a Test Frame In order to test the creation of frames and get an approximation of the time required to render each frame a test frame was rendered on-premise using PolyRay. The output of the rendering process is shown below. The test frame contained 23,629 primitive shapes, most of which are the spheres and cylinders that are used for the 11,800 or so pins in the pin board. The 1280x720 image contains 921,600 pixels, but as anti-aliasing was used the number of rays that were calculated was 4,235,777, with 3,478,754,073 object boundaries checked. The test frame of the pin board with the depth field image applied is shown below. The tracing time for the test frame was 4 minutes 27 seconds, which means rendering the2,000 frames in the animation would take over 148 hours, or a little over 6 days. Although this is much faster that an old 486, waiting almost a week to see the results of an animation would make it challenging for animators to create, view, and refine their animations. It would be much better if the animation could be rendered in less than one hour. Windows Azure Worker Roles The cost of creating an on-premise render farm to render animations increases in proportion to the number of servers. The table below shows the cost of servers for creating a render farm, assuming a cost of $500 per server. Number of Servers Cost 1 $500 16 $8,000 256 $128,000   As well as the cost of the servers, there would be additional costs for networking, racks etc. Hosting an environment of 256 servers on-premise would require a server room with cooling, and some pretty hefty power cabling. The Windows Azure compute services provide worker roles, which are ideal for performing processor intensive compute tasks. With the scalability available in Windows Azure a job that takes 256 hours to complete could be perfumed using different numbers of worker roles. The time and cost of using 1, 16 or 256 worker roles is shown below. Number of Worker Roles Render Time Cost 1 256 hours $30.72 16 16 hours $30.72 256 1 hour $30.72   Using worker roles in Windows Azure provides the same cost for the 256 hour job, irrespective of the number of worker roles used. Provided the compute task can be broken down into many small units, and the worker role compute power can be used effectively, it makes sense to scale the application so that the task is completed quickly, making the results available in a timely fashion. The task of rendering 2,000 frames in an animation is one that can easily be broken down into 2,000 individual pieces, which can be performed by a number of worker roles. Creating a Render Farm in Windows Azure The architecture of the render farm is shown in the following diagram. The render farm is a hybrid application with the following components: ·         On-Premise o   Windows Kinect – Used combined with the Kinect Explorer to create a stream of depth images. o   Animation Creator – This application uses the depth images from the Kinect sensor to create scene description files for PolyRay. These files are then uploaded to the jobs blob container, and job messages added to the jobs queue. o   Process Monitor – This application queries the role instance lifecycle table and displays statistics about the render farm environment and render process. o   Image Downloader – This application polls the image queue and downloads the rendered animation files once they are complete. ·         Windows Azure o   Azure Storage – Queues and blobs are used for the scene description files and completed frames. A table is used to store the statistics about the rendering environment.   The architecture of each worker role is shown below.   The worker role is configured to use local storage, which provides file storage on the worker role instance that can be use by the applications to render the image and transform the format of the image. The service definition for the worker role with the local storage configuration highlighted is shown below. <?xml version="1.0" encoding="utf-8"?> <ServiceDefinition name="CloudRay" >   <WorkerRole name="CloudRayWorkerRole" vmsize="Small">     <Imports>     </Imports>     <ConfigurationSettings>       <Setting name="DataConnectionString" />     </ConfigurationSettings>     <LocalResources>       <LocalStorage name="RayFolder" cleanOnRoleRecycle="true" />     </LocalResources>   </WorkerRole> </ServiceDefinition>     The two executable programs, PolyRay.exe and DTA.exe are included in the Azure project, with Copy Always set as the property. PolyRay will take the scene description file and render it to a Truevision TGA file. As the TGA format has not seen much use since the mid 90’s it is converted to a JPG image using Dave's Targa Animator, another shareware application from the 90’s. Each worker roll will use the following process to render the animation frames. 1.       The worker process polls the job queue, if a job is available the scene description file is downloaded from blob storage to local storage. 2.       PolyRay.exe is started in a process with the appropriate command line arguments to render the image as a TGA file. 3.       DTA.exe is started in a process with the appropriate command line arguments convert the TGA file to a JPG file. 4.       The JPG file is uploaded from local storage to the images blob container. 5.       A message is placed on the images queue to indicate a new image is available for download. 6.       The job message is deleted from the job queue. 7.       The role instance lifecycle table is updated with statistics on the number of frames rendered by the worker role instance, and the CPU time used. The code for this is shown below. public override void Run() {     // Set environment variables     string polyRayPath = Path.Combine(Environment.GetEnvironmentVariable("RoleRoot"), PolyRayLocation);     string dtaPath = Path.Combine(Environment.GetEnvironmentVariable("RoleRoot"), DTALocation);       LocalResource rayStorage = RoleEnvironment.GetLocalResource("RayFolder");     string localStorageRootPath = rayStorage.RootPath;       JobQueue jobQueue = new JobQueue("renderjobs");     JobQueue downloadQueue = new JobQueue("renderimagedownloadjobs");     CloudRayBlob sceneBlob = new CloudRayBlob("scenes");     CloudRayBlob imageBlob = new CloudRayBlob("images");     RoleLifecycleDataSource roleLifecycleDataSource = new RoleLifecycleDataSource();       Frames = 0;       while (true)     {         // Get the render job from the queue         CloudQueueMessage jobMsg = jobQueue.Get();           if (jobMsg != null)         {             // Get the file details             string sceneFile = jobMsg.AsString;             string tgaFile = sceneFile.Replace(".pi", ".tga");             string jpgFile = sceneFile.Replace(".pi", ".jpg");               string sceneFilePath = Path.Combine(localStorageRootPath, sceneFile);             string tgaFilePath = Path.Combine(localStorageRootPath, tgaFile);             string jpgFilePath = Path.Combine(localStorageRootPath, jpgFile);               // Copy the scene file to local storage             sceneBlob.DownloadFile(sceneFilePath);               // Run the ray tracer.             string polyrayArguments =                 string.Format("\"{0}\" -o \"{1}\" -a 2", sceneFilePath, tgaFilePath);             Process polyRayProcess = new Process();             polyRayProcess.StartInfo.FileName =                 Path.Combine(Environment.GetEnvironmentVariable("RoleRoot"), polyRayPath);             polyRayProcess.StartInfo.Arguments = polyrayArguments;             polyRayProcess.Start();             polyRayProcess.WaitForExit();               // Convert the image             string dtaArguments =                 string.Format(" {0} /FJ /P{1}", tgaFilePath, Path.GetDirectoryName (jpgFilePath));             Process dtaProcess = new Process();             dtaProcess.StartInfo.FileName =                 Path.Combine(Environment.GetEnvironmentVariable("RoleRoot"), dtaPath);             dtaProcess.StartInfo.Arguments = dtaArguments;             dtaProcess.Start();             dtaProcess.WaitForExit();               // Upload the image to blob storage             imageBlob.UploadFile(jpgFilePath);               // Add a download job.             downloadQueue.Add(jpgFile);               // Delete the render job message             jobQueue.Delete(jobMsg);               Frames++;         }         else         {             Thread.Sleep(1000);         }           // Log the worker role activity.         roleLifecycleDataSource.Alive             ("CloudRayWorker", RoleLifecycleDataSource.RoleLifecycleId, Frames);     } }     Monitoring Worker Role Instance Lifecycle In order to get more accurate statistics about the lifecycle of the worker role instances used to render the animation data was tracked in an Azure storage table. The following class was used to track the worker role lifecycles in Azure storage.   public class RoleLifecycle : TableServiceEntity {     public string ServerName { get; set; }     public string Status { get; set; }     public DateTime StartTime { get; set; }     public DateTime EndTime { get; set; }     public long SecondsRunning { get; set; }     public DateTime LastActiveTime { get; set; }     public int Frames { get; set; }     public string Comment { get; set; }       public RoleLifecycle()     {     }       public RoleLifecycle(string roleName)     {         PartitionKey = roleName;         RowKey = Utils.GetAscendingRowKey();         Status = "Started";         StartTime = DateTime.UtcNow;         LastActiveTime = StartTime;         EndTime = StartTime;         SecondsRunning = 0;         Frames = 0;     } }     A new instance of this class is created and added to the storage table when the role starts. It is then updated each time the worker renders a frame to record the total number of frames rendered and the total processing time. These statistics are used be the monitoring application to determine the effectiveness of use of resources in the render farm. Rendering the Animation The Azure solution was deployed to Windows Azure with the service configuration set to 16 worker role instances. This allows for the application to be tested in the cloud environment, and the performance of the application determined. When I demo the application at conferences and user groups I often start with 16 instances, and then scale up the application to the full 256 instances. The configuration to run 16 instances is shown below. <?xml version="1.0" encoding="utf-8"?> <ServiceConfiguration serviceName="CloudRay" xmlns="http://schemas.microsoft.com/ServiceHosting/2008/10/ServiceConfiguration" osFamily="1" osVersion="*">   <Role name="CloudRayWorkerRole">     <Instances count="16" />     <ConfigurationSettings>       <Setting name="DataConnectionString"         value="DefaultEndpointsProtocol=https;AccountName=cloudraydata;AccountKey=..." />     </ConfigurationSettings>   </Role> </ServiceConfiguration>     About six minutes after deploying the application the first worker roles become active and start to render the first frames of the animation. The CloudRay Monitor application displays an icon for each worker role instance, with a number indicating the number of frames that the worker role has rendered. The statistics on the left show the number of active worker roles and statistics about the render process. The render time is the time since the first worker role became active; the CPU time is the total amount of processing time used by all worker role instances to render the frames.   Five minutes after the first worker role became active the last of the 16 worker roles activated. By this time the first seven worker roles had each rendered one frame of the animation.   With 16 worker roles u and running it can be seen that one hour and 45 minutes CPU time has been used to render 32 frames with a render time of just under 10 minutes.     At this rate it would take over 10 hours to render the 2,000 frames of the full animation. In order to complete the animation in under an hour more processing power will be required. Scaling the render farm from 16 instances to 256 instances is easy using the new management portal. The slider is set to 256 instances, and the configuration saved. We do not need to re-deploy the application, and the 16 instances that are up and running will not be affected. Alternatively, the configuration file for the Azure service could be modified to specify 256 instances.   <?xml version="1.0" encoding="utf-8"?> <ServiceConfiguration serviceName="CloudRay" xmlns="http://schemas.microsoft.com/ServiceHosting/2008/10/ServiceConfiguration" osFamily="1" osVersion="*">   <Role name="CloudRayWorkerRole">     <Instances count="256" />     <ConfigurationSettings>       <Setting name="DataConnectionString"         value="DefaultEndpointsProtocol=https;AccountName=cloudraydata;AccountKey=..." />     </ConfigurationSettings>   </Role> </ServiceConfiguration>     Six minutes after the new configuration has been applied 75 new worker roles have activated and are processing their first frames.   Five minutes later the full configuration of 256 worker roles is up and running. We can see that the average rate of frame rendering has increased from 3 to 12 frames per minute, and that over 17 hours of CPU time has been utilized in 23 minutes. In this test the time to provision 140 worker roles was about 11 minutes, which works out at about one every five seconds.   We are now half way through the rendering, with 1,000 frames complete. This has utilized just under three days of CPU time in a little over 35 minutes.   The animation is now complete, with 2,000 frames rendered in a little over 52 minutes. The CPU time used by the 256 worker roles is 6 days, 7 hours and 22 minutes with an average frame rate of 38 frames per minute. The rendering of the last 1,000 frames took 16 minutes 27 seconds, which works out at a rendering rate of 60 frames per minute. The frame counts in the server instances indicate that the use of a queue to distribute the workload has been very effective in distributing the load across the 256 worker role instances. The first 16 instances that were deployed first have rendered between 11 and 13 frames each, whilst the 240 instances that were added when the application was scaled have rendered between 6 and 9 frames each.   Completed Animation I’ve uploaded the completed animation to YouTube, a low resolution preview is shown below. Pin Board Animation Created using Windows Kinect and 256 Windows Azure Worker Roles   The animation can be viewed in 1280x720 resolution at the following link: http://www.youtube.com/watch?v=n5jy6bvSxWc Effective Use of Resources According to the CloudRay monitor statistics the animation took 6 days, 7 hours and 22 minutes CPU to render, this works out at 152 hours of compute time, rounded up to the nearest hour. As the usage for the worker role instances are billed for the full hour, it may have been possible to render the animation using fewer than 256 worker roles. When deciding the optimal usage of resources, the time required to provision and start the worker roles must also be considered. In the demo I started with 16 worker roles, and then scaled the application to 256 worker roles. It would have been more optimal to start the application with maybe 200 worker roles, and utilized the full hour that I was being billed for. This would, however, have prevented showing the ease of scalability of the application. The new management portal displays the CPU usage across the worker roles in the deployment. The average CPU usage across all instances is 93.27%, with over 99% used when all the instances are up and running. This shows that the worker role resources are being used very effectively. Grid Computing Scenarios Although I am using this scenario for a hobby project, there are many scenarios where a large amount of compute power is required for a short period of time. Windows Azure provides a great platform for developing these types of grid computing applications, and can work out very cost effective. ·         Windows Azure can provide massive compute power, on demand, in a matter of minutes. ·         The use of queues to manage the load balancing of jobs between role instances is a simple and effective solution. ·         Using a cloud-computing platform like Windows Azure allows proof-of-concept scenarios to be tested and evaluated on a very low budget. ·         No charges for inbound data transfer makes the uploading of large data sets to Windows Azure Storage services cost effective. (Transaction charges still apply.) Tips for using Windows Azure for Grid Computing Scenarios I found the implementation of a render farm using Windows Azure a fairly simple scenario to implement. I was impressed by ease of scalability that Azure provides, and by the short time that the application took to scale from 16 to 256 worker role instances. In this case it was around 13 minutes, in other tests it took between 10 and 20 minutes. The following tips may be useful when implementing a grid computing project in Windows Azure. ·         Using an Azure Storage queue to load-balance the units of work across multiple worker roles is simple and very effective. The design I have used in this scenario could easily scale to many thousands of worker role instances. ·         Windows Azure accounts are typically limited to 20 cores. If you need to use more than this, a call to support and a credit card check will be required. ·         Be aware of how the billing model works. You will be charged for worker role instances for the full clock our in which the instance is deployed. Schedule the workload to start just after the clock hour has started. ·         Monitor the utilization of the resources you are provisioning, ensure that you are not paying for worker roles that are idle. ·         If you are deploying third party applications to worker roles, you may well run into licensing issues. Purchasing software licenses on a per-processor basis when using hundreds of processors for a short time period would not be cost effective. ·         Third party software may also require installation onto the worker roles, which can be accomplished using start-up tasks. Bear in mind that adding a startup task and possible re-boot will add to the time required for the worker role instance to start and activate. An alternative may be to use a prepared VM and use VM roles. ·         Consider using the Windows Azure Autoscaling Application Block (WASABi) to autoscale the worker roles in your application. When using a large number of worker roles, the utilization must be carefully monitored, if the scaling algorithms are not optimal it could get very expensive!

  Read the article

• Using R to Analyze G1GC Log Files

  - by user12620111

  Using R to Analyze G1GC Log Files body, td { font-family: sans-serif; background-color: white; font-size: 12px; margin: 8px; } tt, code, pre { font-family: 'DejaVu Sans Mono', 'Droid Sans Mono', 'Lucida Console', Consolas, Monaco, monospace; } h1 { font-size:2.2em; } h2 { font-size:1.8em; } h3 { font-size:1.4em; } h4 { font-size:1.0em; } h5 { font-size:0.9em; } h6 { font-size:0.8em; } a:visited { color: rgb(50%, 0%, 50%); } pre { margin-top: 0; max-width: 95%; border: 1px solid #ccc; white-space: pre-wrap; } pre code { display: block; padding: 0.5em; } code.r, code.cpp { background-color: #F8F8F8; } table, td, th { border: none; } blockquote { color:#666666; margin:0; padding-left: 1em; border-left: 0.5em #EEE solid; } hr { height: 0px; border-bottom: none; border-top-width: thin; border-top-style: dotted; border-top-color: #999999; } @media print { * { background: transparent !important; color: black !important; filter:none !important; -ms-filter: none !important; } body { font-size:12pt; max-width:100%; } a, a:visited { text-decoration: underline; } hr { visibility: hidden; page-break-before: always; } pre, blockquote { padding-right: 1em; page-break-inside: avoid; } tr, img { page-break-inside: avoid; } img { max-width: 100% !important; } @page :left { margin: 15mm 20mm 15mm 10mm; } @page :right { margin: 15mm 10mm 15mm 20mm; } p, h2, h3 { orphans: 3; widows: 3; } h2, h3 { page-break-after: avoid; } } pre .operator, pre .paren { color: rgb(104, 118, 135) } pre .literal { color: rgb(88, 72, 246) } pre .number { color: rgb(0, 0, 205); } pre .comment { color: rgb(76, 136, 107); } pre .keyword { color: rgb(0, 0, 255); } pre .identifier { color: rgb(0, 0, 0); } pre .string { color: rgb(3, 106, 7); } var hljs=new function(){function m(p){return p.replace(/&/gm,"&").replace(/"}while(y.length||w.length){var v=u().splice(0,1)[0];z+=m(x.substr(q,v.offset-q));q=v.offset;if(v.event=="start"){z+=t(v.node);s.push(v.node)}else{if(v.event=="stop"){var p,r=s.length;do{r--;p=s[r];z+=("")}while(p!=v.node);s.splice(r,1);while(r'+M[0]+""}else{r+=M[0]}O=P.lR.lastIndex;M=P.lR.exec(L)}return r+L.substr(O,L.length-O)}function J(L,M){if(M.sL&&e[M.sL]){var r=d(M.sL,L);x+=r.keyword_count;return r.value}else{return F(L,M)}}function I(M,r){var L=M.cN?'':"";if(M.rB){y+=L;M.buffer=""}else{if(M.eB){y+=m(r)+L;M.buffer=""}else{y+=L;M.buffer=r}}D.push(M);A+=M.r}function G(N,M,Q){var R=D[D.length-1];if(Q){y+=J(R.buffer+N,R);return false}var P=q(M,R);if(P){y+=J(R.buffer+N,R);I(P,M);return P.rB}var L=v(D.length-1,M);if(L){var O=R.cN?"":"";if(R.rE){y+=J(R.buffer+N,R)+O}else{if(R.eE){y+=J(R.buffer+N,R)+O+m(M)}else{y+=J(R.buffer+N+M,R)+O}}while(L1){O=D[D.length-2].cN?"":"";y+=O;L--;D.length--}var r=D[D.length-1];D.length--;D[D.length-1].buffer="";if(r.starts){I(r.starts,"")}return R.rE}if(w(M,R)){throw"Illegal"}}var E=e[B];var D=[E.dM];var A=0;var x=0;var y="";try{var s,u=0;E.dM.buffer="";do{s=p(C,u);var t=G(s[0],s[1],s[2]);u+=s[0].length;if(!t){u+=s[1].length}}while(!s[2]);if(D.length1){throw"Illegal"}return{r:A,keyword_count:x,value:y}}catch(H){if(H=="Illegal"){return{r:0,keyword_count:0,value:m(C)}}else{throw H}}}function g(t){var p={keyword_count:0,r:0,value:m(t)};var r=p;for(var q in e){if(!e.hasOwnProperty(q)){continue}var s=d(q,t);s.language=q;if(s.keyword_count+s.rr.keyword_count+r.r){r=s}if(s.keyword_count+s.rp.keyword_count+p.r){r=p;p=s}}if(r.language){p.second_best=r}return p}function i(r,q,p){if(q){r=r.replace(/^((]+|\t)+)/gm,function(t,w,v,u){return w.replace(/\t/g,q)})}if(p){r=r.replace(/\n/g,"")}return r}function n(t,w,r){var x=h(t,r);var v=a(t);var y,s;if(v){y=d(v,x)}else{return}var q=c(t);if(q.length){s=document.createElement("pre");s.innerHTML=y.value;y.value=k(q,c(s),x)}y.value=i(y.value,w,r);var u=t.className;if(!u.match("(\\s|^)(language-)?"+v+"(\\s|$)")){u=u?(u+" "+v):v}if(/MSIE [678]/.test(navigator.userAgent)&&t.tagName=="CODE"&&t.parentNode.tagName=="PRE"){s=t.parentNode;var p=document.createElement("div");p.innerHTML=""+y.value+"";t=p.firstChild.firstChild;p.firstChild.cN=s.cN;s.parentNode.replaceChild(p.firstChild,s)}else{t.innerHTML=y.value}t.className=u;t.result={language:v,kw:y.keyword_count,re:y.r};if(y.second_best){t.second_best={language:y.second_best.language,kw:y.second_best.keyword_count,re:y.second_best.r}}}function o(){if(o.called){return}o.called=true;var r=document.getElementsByTagName("pre");for(var p=0;p|=||=||=|\\?|\\[|\\{|\\(|\\^|\\^=|\\||\\|=|\\|\\||~";this.ER="(?![\\s\\S])";this.BE={b:"\\\\.",r:0};this.ASM={cN:"string",b:"'",e:"'",i:"\\n",c:[this.BE],r:0};this.QSM={cN:"string",b:'"',e:'"',i:"\\n",c:[this.BE],r:0};this.CLCM={cN:"comment",b:"//",e:"$"};this.CBLCLM={cN:"comment",b:"/\\*",e:"\\*/"};this.HCM={cN:"comment",b:"#",e:"$"};this.NM={cN:"number",b:this.NR,r:0};this.CNM={cN:"number",b:this.CNR,r:0};this.BNM={cN:"number",b:this.BNR,r:0};this.inherit=function(r,s){var p={};for(var q in r){p[q]=r[q]}if(s){for(var q in s){p[q]=s[q]}}return p}}();hljs.LANGUAGES.cpp=function(){var a={keyword:{"false":1,"int":1,"float":1,"while":1,"private":1,"char":1,"catch":1,"export":1,virtual:1,operator:2,sizeof:2,dynamic_cast:2,typedef:2,const_cast:2,"const":1,struct:1,"for":1,static_cast:2,union:1,namespace:1,unsigned:1,"long":1,"throw":1,"volatile":2,"static":1,"protected":1,bool:1,template:1,mutable:1,"if":1,"public":1,friend:2,"do":1,"return":1,"goto":1,auto:1,"void":2,"enum":1,"else":1,"break":1,"new":1,extern:1,using:1,"true":1,"class":1,asm:1,"case":1,typeid:1,"short":1,reinterpret_cast:2,"default":1,"double":1,register:1,explicit:1,signed:1,typename:1,"try":1,"this":1,"switch":1,"continue":1,wchar_t:1,inline:1,"delete":1,alignof:1,char16_t:1,char32_t:1,constexpr:1,decltype:1,noexcept:1,nullptr:1,static_assert:1,thread_local:1,restrict:1,_Bool:1,complex:1},built_in:{std:1,string:1,cin:1,cout:1,cerr:1,clog:1,stringstream:1,istringstream:1,ostringstream:1,auto_ptr:1,deque:1,list:1,queue:1,stack:1,vector:1,map:1,set:1,bitset:1,multiset:1,multimap:1,unordered_set:1,unordered_map:1,unordered_multiset:1,unordered_multimap:1,array:1,shared_ptr:1}};return{dM:{k:a,i:"",k:a,r:10,c:["self"]}]}}}();hljs.LANGUAGES.r={dM:{c:[hljs.HCM,{cN:"number",b:"\\b0[xX][0-9a-fA-F]+[Li]?\\b",e:hljs.IMMEDIATE_RE,r:0},{cN:"number",b:"\\b\\d+(?:[eE][+\\-]?\\d*)?L\\b",e:hljs.IMMEDIATE_RE,r:0},{cN:"number",b:"\\b\\d+\\.(?!\\d)(?:i\\b)?",e:hljs.IMMEDIATE_RE,r:1},{cN:"number",b:"\\b\\d+(?:\\.\\d*)?(?:[eE][+\\-]?\\d*)?i?\\b",e:hljs.IMMEDIATE_RE,r:0},{cN:"number",b:"\\.\\d+(?:[eE][+\\-]?\\d*)?i?\\b",e:hljs.IMMEDIATE_RE,r:1},{cN:"keyword",b:"(?:tryCatch|library|setGeneric|setGroupGeneric)\\b",e:hljs.IMMEDIATE_RE,r:10},{cN:"keyword",b:"\\.\\.\\.",e:hljs.IMMEDIATE_RE,r:10},{cN:"keyword",b:"\\.\\.\\d+(?![\\w.])",e:hljs.IMMEDIATE_RE,r:10},{cN:"keyword",b:"\\b(?:function)",e:hljs.IMMEDIATE_RE,r:2},{cN:"keyword",b:"(?:if|in|break|next|repeat|else|for|return|switch|while|try|stop|warning|require|attach|detach|source|setMethod|setClass)\\b",e:hljs.IMMEDIATE_RE,r:1},{cN:"literal",b:"(?:NA|NA_integer_|NA_real_|NA_character_|NA_complex_)\\b",e:hljs.IMMEDIATE_RE,r:10},{cN:"literal",b:"(?:NULL|TRUE|FALSE|T|F|Inf|NaN)\\b",e:hljs.IMMEDIATE_RE,r:1},{cN:"identifier",b:"[a-zA-Z.][a-zA-Z0-9._]*\\b",e:hljs.IMMEDIATE_RE,r:0},{cN:"operator",b:"|=||   Using R to Analyze G1GC Log Files   Using R to Analyze G1GC Log Files Introduction Working in Oracle Platform Integration gives an engineer opportunities to work on a wide array of technologies. My team’s goal is to make Oracle applications run best on the Solaris/SPARC platform. When looking for bottlenecks in a modern applications, one needs to be aware of not only how the CPUs and operating system are executing, but also network, storage, and in some cases, the Java Virtual Machine. I was recently presented with about 1.5 GB of Java Garbage First Garbage Collector log file data. If you’re not familiar with the subject, you might want to review Garbage First Garbage Collector Tuning by Monica Beckwith. The customer had been running Java HotSpot 1.6.0_31 to host a web application server. I was told that the Solaris/SPARC server was running a Java process launched using a commmand line that included the following flags: -d64 -Xms9g -Xmx9g -XX:+UseG1GC -XX:MaxGCPauseMillis=200 -XX:InitiatingHeapOccupancyPercent=80 -XX:PermSize=256m -XX:MaxPermSize=256m -XX:+PrintGC -XX:+PrintGCTimeStamps -XX:+PrintHeapAtGC -XX:+PrintGCDateStamps -XX:+PrintFlagsFinal -XX:+DisableExplicitGC -XX:+UnlockExperimentalVMOptions -XX:ParallelGCThreads=8 Several sources on the internet indicate that if I were to print out the 1.5 GB of log files, it would require enough paper to fill the bed of a pick up truck. Of course, it would be fruitless to try to scan the log files by hand. Tools will be required to summarize the contents of the log files. Others have encountered large Java garbage collection log files. There are existing tools to analyze the log files: IBM’s GC toolkit The chewiebug GCViewer gchisto HPjmeter Instead of using one of the other tools listed, I decide to parse the log files with standard Unix tools, and analyze the data with R. Data Cleansing The log files arrived in two different formats. I guess that the difference is that one set of log files was generated using a more verbose option, maybe -XX:+PrintHeapAtGC, and the other set of log files was generated without that option. Format 1 In some of the log files, the log files with the less verbose format, a single trace, i.e. the report of a singe garbage collection event, looks like this: {Heap before GC invocations=12280 (full 61): garbage-first heap total 9437184K, used 7499918K [0xfffffffd00000000, 0xffffffff40000000, 0xffffffff40000000) region size 4096K, 1 young (4096K), 0 survivors (0K) compacting perm gen total 262144K, used 144077K [0xffffffff40000000, 0xffffffff50000000, 0xffffffff50000000) the space 262144K, 54% used [0xffffffff40000000, 0xffffffff48cb3758, 0xffffffff48cb3800, 0xffffffff50000000) No shared spaces configured. 2014-05-14T07:24:00.988-0700: 60586.353: [GC pause (young) 7324M->7320M(9216M), 0.1567265 secs] Heap after GC invocations=12281 (full 61): garbage-first heap total 9437184K, used 7496533K [0xfffffffd00000000, 0xffffffff40000000, 0xffffffff40000000) region size 4096K, 0 young (0K), 0 survivors (0K) compacting perm gen total 262144K, used 144077K [0xffffffff40000000, 0xffffffff50000000, 0xffffffff50000000) the space 262144K, 54% used [0xffffffff40000000, 0xffffffff48cb3758, 0xffffffff48cb3800, 0xffffffff50000000) No shared spaces configured. } A simple grep can be used to extract a summary: $ grep "\[ GC pause (young" g1gc.log 2014-05-13T13:24:35.091-0700: 3.109: [GC pause (young) 20M->5029K(9216M), 0.0146328 secs] 2014-05-13T13:24:35.440-0700: 3.459: [GC pause (young) 9125K->6077K(9216M), 0.0086723 secs] 2014-05-13T13:24:37.581-0700: 5.599: [GC pause (young) 25M->8470K(9216M), 0.0203820 secs] 2014-05-13T13:24:42.686-0700: 10.704: [GC pause (young) 44M->15M(9216M), 0.0288848 secs] 2014-05-13T13:24:48.941-0700: 16.958: [GC pause (young) 51M->20M(9216M), 0.0491244 secs] 2014-05-13T13:24:56.049-0700: 24.066: [GC pause (young) 92M->26M(9216M), 0.0525368 secs] 2014-05-13T13:25:34.368-0700: 62.383: [GC pause (young) 602M->68M(9216M), 0.1721173 secs] But that format wasn't easily read into R, so I needed to be a bit more tricky. I used the following Unix command to create a summary file that was easy for R to read. $ echo "SecondsSinceLaunch BeforeSize AfterSize TotalSize RealTime" $ grep "\[GC pause (young" g1gc.log | grep -v mark | sed -e 's/[A-SU-z\(\),]/ /g' -e 's/->/ /' -e 's/: / /g' | more SecondsSinceLaunch BeforeSize AfterSize TotalSize RealTime 2014-05-13T13:24:35.091-0700 3.109 20 5029 9216 0.0146328 2014-05-13T13:24:35.440-0700 3.459 9125 6077 9216 0.0086723 2014-05-13T13:24:37.581-0700 5.599 25 8470 9216 0.0203820 2014-05-13T13:24:42.686-0700 10.704 44 15 9216 0.0288848 2014-05-13T13:24:48.941-0700 16.958 51 20 9216 0.0491244 2014-05-13T13:24:56.049-0700 24.066 92 26 9216 0.0525368 2014-05-13T13:25:34.368-0700 62.383 602 68 9216 0.1721173 Format 2 In some of the log files, the log files with the more verbose format, a single trace, i.e. the report of a singe garbage collection event, was more complicated than Format 1. Here is a text file with an example of a single G1GC trace in the second format. As you can see, it is quite complicated. It is nice that there is so much information available, but the level of detail can be overwhelming. I wrote this awk script (download) to summarize each trace on a single line. #!/usr/bin/env awk -f BEGIN { printf("SecondsSinceLaunch IncrementalCount FullCount UserTime SysTime RealTime BeforeSize AfterSize TotalSize\n") } ###################### # Save count data from lines that are at the start of each G1GC trace. # Each trace starts out like this: # {Heap before GC invocations=14 (full 0): # garbage-first heap total 9437184K, used 325496K [0xfffffffd00000000, 0xffffffff40000000, 0xffffffff40000000) ###################### /{Heap.*full/{ gsub ( "\\)" , "" ); nf=split($0,a,"="); split(a[2],b," "); getline; if ( match($0, "first") ) { G1GC=1; IncrementalCount=b[1]; FullCount=substr( b[3], 1, length(b[3])-1 ); } else { G1GC=0; } } ###################### # Pull out time stamps that are in lines with this format: # 2014-05-12T14:02:06.025-0700: 94.312: [GC pause (young), 0.08870154 secs] ###################### /GC pause/ { DateTime=$1; SecondsSinceLaunch=substr($2, 1, length($2)-1); } ###################### # Heap sizes are in lines that look like this: # [ 4842M->4838M(9216M)] ###################### /\[ .*]$/ { gsub ( "\\[" , "" ); gsub ( "\ \]" , "" ); gsub ( "->" , " " ); gsub ( "\\( " , " " ); gsub ( "\ \)" , " " ); split($0,a," "); if ( split(a[1],b,"M") > 1 ) {BeforeSize=b[1]*1024;} if ( split(a[1],b,"K") > 1 ) {BeforeSize=b[1];} if ( split(a[2],b,"M") > 1 ) {AfterSize=b[1]*1024;} if ( split(a[2],b,"K") > 1 ) {AfterSize=b[1];} if ( split(a[3],b,"M") > 1 ) {TotalSize=b[1]*1024;} if ( split(a[3],b,"K") > 1 ) {TotalSize=b[1];} } ###################### # Emit an output line when you find input that looks like this: # [Times: user=1.41 sys=0.08, real=0.24 secs] ###################### /\[Times/ { if (G1GC==1) { gsub ( "," , "" ); split($2,a,"="); UserTime=a[2]; split($3,a,"="); SysTime=a[2]; split($4,a,"="); RealTime=a[2]; print DateTime,SecondsSinceLaunch,IncrementalCount,FullCount,UserTime,SysTime,RealTime,BeforeSize,AfterSize,TotalSize; G1GC=0; } } The resulting summary is about 25X smaller that the original file, but still difficult for a human to digest. SecondsSinceLaunch IncrementalCount FullCount UserTime SysTime RealTime BeforeSize AfterSize TotalSize ... 2014-05-12T18:36:34.669-0700: 3985.744 561 0 0.57 0.06 0.16 1724416 1720320 9437184 2014-05-12T18:36:34.839-0700: 3985.914 562 0 0.51 0.06 0.19 1724416 1720320 9437184 2014-05-12T18:36:35.069-0700: 3986.144 563 0 0.60 0.04 0.27 1724416 1721344 9437184 2014-05-12T18:36:35.354-0700: 3986.429 564 0 0.33 0.04 0.09 1725440 1722368 9437184 2014-05-12T18:36:35.545-0700: 3986.620 565 0 0.58 0.04 0.17 1726464 1722368 9437184 2014-05-12T18:36:35.726-0700: 3986.801 566 0 0.43 0.05 0.12 1726464 1722368 9437184 2014-05-12T18:36:35.856-0700: 3986.930 567 0 0.30 0.04 0.07 1726464 1723392 9437184 2014-05-12T18:36:35.947-0700: 3987.023 568 0 0.61 0.04 0.26 1727488 1723392 9437184 2014-05-12T18:36:36.228-0700: 3987.302 569 0 0.46 0.04 0.16 1731584 1724416 9437184 Reading the Data into R Once the GC log data had been cleansed, either by processing the first format with the shell script, or by processing the second format with the awk script, it was easy to read the data into R. g1gc.df = read.csv("summary.txt", row.names = NULL, stringsAsFactors=FALSE,sep="") str(g1gc.df) ## 'data.frame': 8307 obs. of 10 variables: ## $ row.names : chr "2014-05-12T14:00:32.868-0700:" "2014-05-12T14:00:33.179-0700:" "2014-05-12T14:00:33.677-0700:" "2014-05-12T14:00:35.538-0700:" ... ## $ SecondsSinceLaunch: num 1.16 1.47 1.97 3.83 6.1 ... ## $ IncrementalCount : int 0 1 2 3 4 5 6 7 8 9 ... ## $ FullCount : int 0 0 0 0 0 0 0 0 0 0 ... ## $ UserTime : num 0.11 0.05 0.04 0.21 0.08 0.26 0.31 0.33 0.34 0.56 ... ## $ SysTime : num 0.04 0.01 0.01 0.05 0.01 0.06 0.07 0.06 0.07 0.09 ... ## $ RealTime : num 0.02 0.02 0.01 0.04 0.02 0.04 0.05 0.04 0.04 0.06 ... ## $ BeforeSize : int 8192 5496 5768 22528 24576 43008 34816 53248 55296 93184 ... ## $ AfterSize : int 1400 1672 2557 4907 7072 14336 16384 18432 19456 21504 ... ## $ TotalSize : int 9437184 9437184 9437184 9437184 9437184 9437184 9437184 9437184 9437184 9437184 ... head(g1gc.df) ## row.names SecondsSinceLaunch IncrementalCount ## 1 2014-05-12T14:00:32.868-0700: 1.161 0 ## 2 2014-05-12T14:00:33.179-0700: 1.472 1 ## 3 2014-05-12T14:00:33.677-0700: 1.969 2 ## 4 2014-05-12T14:00:35.538-0700: 3.830 3 ## 5 2014-05-12T14:00:37.811-0700: 6.103 4 ## 6 2014-05-12T14:00:41.428-0700: 9.720 5 ## FullCount UserTime SysTime RealTime BeforeSize AfterSize TotalSize ## 1 0 0.11 0.04 0.02 8192 1400 9437184 ## 2 0 0.05 0.01 0.02 5496 1672 9437184 ## 3 0 0.04 0.01 0.01 5768 2557 9437184 ## 4 0 0.21 0.05 0.04 22528 4907 9437184 ## 5 0 0.08 0.01 0.02 24576 7072 9437184 ## 6 0 0.26 0.06 0.04 43008 14336 9437184 Basic Statistics Once the data has been read into R, simple statistics are very easy to generate. All of the numbers from high school statistics are available via simple commands. For example, generate a summary of every column: summary(g1gc.df) ## row.names SecondsSinceLaunch IncrementalCount FullCount ## Length:8307 Min. : 1 Min. : 0 Min. : 0.0 ## Class :character 1st Qu.: 9977 1st Qu.:2048 1st Qu.: 0.0 ## Mode :character Median :12855 Median :4136 Median : 12.0 ## Mean :12527 Mean :4156 Mean : 31.6 ## 3rd Qu.:15758 3rd Qu.:6262 3rd Qu.: 61.0 ## Max. :55484 Max. :8391 Max. :113.0 ## UserTime SysTime RealTime BeforeSize ## Min. :0.040 Min. :0.0000 Min. : 0.0 Min. : 5476 ## 1st Qu.:0.470 1st Qu.:0.0300 1st Qu.: 0.1 1st Qu.:5137920 ## Median :0.620 Median :0.0300 Median : 0.1 Median :6574080 ## Mean :0.751 Mean :0.0355 Mean : 0.3 Mean :5841855 ## 3rd Qu.:0.920 3rd Qu.:0.0400 3rd Qu.: 0.2 3rd Qu.:7084032 ## Max. :3.370 Max. :1.5600 Max. :488.1 Max. :8696832 ## AfterSize TotalSize ## Min. : 1380 Min. :9437184 ## 1st Qu.:5002752 1st Qu.:9437184 ## Median :6559744 Median :9437184 ## Mean :5785454 Mean :9437184 ## 3rd Qu.:7054336 3rd Qu.:9437184 ## Max. :8482816 Max. :9437184 Q: What is the total amount of User CPU time spent in garbage collection? sum(g1gc.df$UserTime) ## [1] 6236 As you can see, less than two hours of CPU time was spent in garbage collection. Is that too much? To find the percentage of time spent in garbage collection, divide the number above by total_elapsed_time*CPU_count. In this case, there are a lot of CPU’s and it turns out the the overall amount of CPU time spent in garbage collection isn’t a problem when viewed in isolation. When calculating rates, i.e. events per unit time, you need to ask yourself if the rate is homogenous across the time period in the log file. Does the log file include spikes of high activity that should be separately analyzed? Averaging in data from nights and weekends with data from business hours may alias problems. If you have a reason to suspect that the garbage collection rates include peaks and valleys that need independent analysis, see the “Time Series” section, below. Q: How much garbage is collected on each pass? The amount of heap space that is recovered per GC pass is surprisingly low: At least one collection didn’t recover any data. (“Min.=0”) 25% of the passes recovered 3MB or less. (“1st Qu.=3072”) Half of the GC passes recovered 4MB or less. (“Median=4096”) The average amount recovered was 56MB. (“Mean=56390”) 75% of the passes recovered 36MB or less. (“3rd Qu.=36860”) At least one pass recovered 2GB. (“Max.=2121000”) g1gc.df$Delta = g1gc.df$BeforeSize - g1gc.df$AfterSize summary(g1gc.df$Delta) ## Min. 1st Qu. Median Mean 3rd Qu. Max. ## 0 3070 4100 56400 36900 2120000 Q: What is the maximum User CPU time for a single collection? The worst garbage collection (“Max.”) is many standard deviations away from the mean. The data appears to be right skewed. summary(g1gc.df$UserTime) ## Min. 1st Qu. Median Mean 3rd Qu. Max. ## 0.040 0.470 0.620 0.751 0.920 3.370 sd(g1gc.df$UserTime) ## [1] 0.3966 Basic Graphics Once the data is in R, it is trivial to plot the data with formats including dot plots, line charts, bar charts (simple, stacked, grouped), pie charts, boxplots, scatter plots histograms, and kernel density plots. Histogram of User CPU Time per Collection I don't think that this graph requires any explanation. hist(g1gc.df$UserTime, main="User CPU Time per Collection", xlab="Seconds", ylab="Frequency") Box plot to identify outliers When the initial data is viewed with a box plot, you can see the one crazy outlier in the real time per GC. Save this data point for future analysis and drop the outlier so that it’s not throwing off our statistics. Now the box plot shows many outliers, which will be examined later, using times series analysis. Notice that the scale of the x-axis changes drastically once the crazy outlier is removed. par(mfrow=c(2,1)) boxplot(g1gc.df$UserTime,g1gc.df$SysTime,g1gc.df$RealTime, main="Box Plot of Time per GC\n(dominated by a crazy outlier)", names=c("usr","sys","elapsed"), xlab="Seconds per GC", ylab="Time (Seconds)", horizontal = TRUE, outcol="red") crazy.outlier.df=g1gc.df[g1gc.df$RealTime > 400,] g1gc.df=g1gc.df[g1gc.df$RealTime < 400,] boxplot(g1gc.df$UserTime,g1gc.df$SysTime,g1gc.df$RealTime, main="Box Plot of Time per GC\n(crazy outlier excluded)", names=c("usr","sys","elapsed"), xlab="Seconds per GC", ylab="Time (Seconds)", horizontal = TRUE, outcol="red") box(which = "outer", lty = "solid") Here is the crazy outlier for future analysis: crazy.outlier.df ## row.names SecondsSinceLaunch IncrementalCount ## 8233 2014-05-12T23:15:43.903-0700: 20741 8316 ## FullCount UserTime SysTime RealTime BeforeSize AfterSize TotalSize ## 8233 112 0.55 0.42 488.1 8381440 8235008 9437184 ## Delta ## 8233 146432 R Time Series Data To analyze the garbage collection as a time series, I’ll use Z’s Ordered Observations (zoo). “zoo is the creator for an S3 class of indexed totally ordered observations which includes irregular time series.” require(zoo) ## Loading required package: zoo ## ## Attaching package: 'zoo' ## ## The following objects are masked from 'package:base': ## ## as.Date, as.Date.numeric head(g1gc.df[,1]) ## [1] "2014-05-12T14:00:32.868-0700:" "2014-05-12T14:00:33.179-0700:" ## [3] "2014-05-12T14:00:33.677-0700:" "2014-05-12T14:00:35.538-0700:" ## [5] "2014-05-12T14:00:37.811-0700:" "2014-05-12T14:00:41.428-0700:" options("digits.secs"=3) times=as.POSIXct( g1gc.df[,1], format="%Y-%m-%dT%H:%M:%OS%z:") g1gc.z = zoo(g1gc.df[,-c(1)], order.by=times) head(g1gc.z) ## SecondsSinceLaunch IncrementalCount FullCount ## 2014-05-12 17:00:32.868 1.161 0 0 ## 2014-05-12 17:00:33.178 1.472 1 0 ## 2014-05-12 17:00:33.677 1.969 2 0 ## 2014-05-12 17:00:35.538 3.830 3 0 ## 2014-05-12 17:00:37.811 6.103 4 0 ## 2014-05-12 17:00:41.427 9.720 5 0 ## UserTime SysTime RealTime BeforeSize AfterSize ## 2014-05-12 17:00:32.868 0.11 0.04 0.02 8192 1400 ## 2014-05-12 17:00:33.178 0.05 0.01 0.02 5496 1672 ## 2014-05-12 17:00:33.677 0.04 0.01 0.01 5768 2557 ## 2014-05-12 17:00:35.538 0.21 0.05 0.04 22528 4907 ## 2014-05-12 17:00:37.811 0.08 0.01 0.02 24576 7072 ## 2014-05-12 17:00:41.427 0.26 0.06 0.04 43008 14336 ## TotalSize Delta ## 2014-05-12 17:00:32.868 9437184 6792 ## 2014-05-12 17:00:33.178 9437184 3824 ## 2014-05-12 17:00:33.677 9437184 3211 ## 2014-05-12 17:00:35.538 9437184 17621 ## 2014-05-12 17:00:37.811 9437184 17504 ## 2014-05-12 17:00:41.427 9437184 28672 Example of Two Benchmark Runs in One Log File The data in the following graph is from a different log file, not the one of primary interest to this article. I’m including this image because it is an example of idle periods followed by busy periods. It would be uninteresting to average the rate of garbage collection over the entire log file period. More interesting would be the rate of garbage collect in the two busy periods. Are they the same or different? Your production data may be similar, for example, bursts when employees return from lunch and idle times on weekend evenings, etc. Once the data is in an R Time Series, you can analyze isolated time windows. Clipping the Time Series data Flashing back to our test case… Viewing the data as a time series is interesting. You can see that the work intensive time period is between 9:00 PM and 3:00 AM. Lets clip the data to the interesting period:     par(mfrow=c(2,1)) plot(g1gc.z$UserTime, type="h", main="User Time per GC\nTime: Complete Log File", xlab="Time of Day", ylab="CPU Seconds per GC", col="#1b9e77") clipped.g1gc.z=window(g1gc.z, start=as.POSIXct("2014-05-12 21:00:00"), end=as.POSIXct("2014-05-13 03:00:00")) plot(clipped.g1gc.z$UserTime, type="h", main="User Time per GC\nTime: Limited to Benchmark Execution", xlab="Time of Day", ylab="CPU Seconds per GC", col="#1b9e77") box(which = "outer", lty = "solid") Cumulative Incremental and Full GC count Here is the cumulative incremental and full GC count. When the line is very steep, it indicates that the GCs are repeating very quickly. Notice that the scale on the Y axis is different for full vs. incremental. plot(clipped.g1gc.z[,c(2:3)], main="Cumulative Incremental and Full GC count", xlab="Time of Day", col="#1b9e77") GC Analysis of Benchmark Execution using Time Series data In the following series of 3 graphs: The “After Size” show the amount of heap space in use after each garbage collection. Many Java objects are still referenced, i.e. alive, during each garbage collection. This may indicate that the application has a memory leak, or may indicate that the application has a very large memory footprint. Typically, an application's memory footprint plateau's in the early stage of execution. One would expect this graph to have a flat top. The steep decline in the heap space may indicate that the application crashed after 2:00. The second graph shows that the outliers in real execution time, discussed above, occur near 2:00. when the Java heap seems to be quite full. The third graph shows that Full GCs are infrequent during the first few hours of execution. The rate of Full GC's, (the slope of the cummulative Full GC line), changes near midnight.   plot(clipped.g1gc.z[,c("AfterSize","RealTime","FullCount")], xlab="Time of Day", col=c("#1b9e77","red","#1b9e77")) GC Analysis of heap recovered Each GC trace includes the amount of heap space in use before and after the individual GC event. During garbage coolection, unreferenced objects are identified, the space holding the unreferenced objects is freed, and thus, the difference in before and after usage indicates how much space has been freed. The following box plot and bar chart both demonstrate the same point - the amount of heap space freed per garbage colloection is surprisingly low. par(mfrow=c(2,1)) boxplot(as.vector(clipped.g1gc.z$Delta), main="Amount of Heap Recovered per GC Pass", xlab="Size in KB", horizontal = TRUE, col="red") hist(as.vector(clipped.g1gc.z$Delta), main="Amount of Heap Recovered per GC Pass", xlab="Size in KB", breaks=100, col="red") box(which = "outer", lty = "solid") This graph is the most interesting. The dark blue area shows how much heap is occupied by referenced Java objects. This represents memory that holds live data. The red fringe at the top shows how much data was recovered after each garbage collection. barplot(clipped.g1gc.z[,c("AfterSize","Delta")], col=c("#7570b3","#e7298a"), xlab="Time of Day", border=NA) legend("topleft", c("Live Objects","Heap Recovered on GC"), fill=c("#7570b3","#e7298a")) box(which = "outer", lty = "solid") When I discuss the data in the log files with the customer, I will ask for an explaination for the large amount of referenced data resident in the Java heap. There are two are posibilities: There is a memory leak and the amount of space required to hold referenced objects will continue to grow, limited only by the maximum heap size. After the maximum heap size is reached, the JVM will throw an “Out of Memory” exception every time that the application tries to allocate a new object. If this is the case, the aplication needs to be debugged to identify why old objects are referenced when they are no longer needed. The application has a legitimate requirement to keep a large amount of data in memory. The customer may want to further increase the maximum heap size. Another possible solution would be to partition the application across multiple cluster nodes, where each node has responsibility for managing a unique subset of the data. Conclusion In conclusion, R is a very powerful tool for the analysis of Java garbage collection log files. The primary difficulty is data cleansing so that information can be read into an R data frame. Once the data has been read into R, a rich set of tools may be used for thorough evaluation.

  Read the article

< Previous Page | 124 125 126 127 128 129 130 131 132 133 134 135  | Next Page >